(1) | They wonder whether the lions will devour the wildebeest. |
The treelets for wonder and whether are given in (2a,b), and the entire tree for (1) is given in (2c).
(2) | a. | ||
b. | |||
c. |
Now consider the indirect questions in (3), which are each introduced by an entire wh-phrase instead of by a complementizer.
(3) | a. | They wonder which wildebeest the lions will devour. | |
b. | They wonder which of the wildebeests the lions will devour. |
Let's adopt the null hypothesis that wonder is associated with the same treelet in (3) as it is in (1)namely, with (2a). Since the sentences in (3) contain no overt complementizer, we must then assume that the CP tree that substitutes into the complement node of the wonder treelet is the projection of a silent complementizer. Assuming such a silent complementizer is nothing new, of course; it is simply analogous to the assumption that the tree for the matrix clause They wonder in (1) and (3) is the projection of a silent counterpart of do. The matrix clause for (3) can now be constructed straightforwardly, and so can part of the indirect question, as shown in (4a,b).
Note: For simplicity, we focus on (3a). "X" is a variable ranging over D and P; that is, XP can be replaced by DP or PP.
(4) | a. | b. |
But there is a difficulty concerning the wh-phrase which wildebeest. On the one hand, the wh-phrase must function as the object of devour, just as it does in (1), because the verb is obligatorily transitive. On the other hand, the wh-phrase precedes the subject of the subordinate clause rather than following the verb. The obvious solution to this dilemma is to say that the wh-phrase originates as the sister of the verb whose object it is and then moves to the available slot in Spec(CP). We can think of movement as allowing a single phrase to fulfill functions in two different places at the same timea form of bilocation, as it were. In particular, assuming wh-movement in (3a) allows us to maintain that devour is a transitive verb regardless of whether it occurs in declarative clauses or (indirect) questions without denying the word order facts. The full structure of (3a) is shown in (5).
(5) |
Incidentally, in case you are worried that the CP projecting from a silent C node is a cheap trick for which there is no independent evidence, it is worth noting that in Middle English, the equivalent of (3a) would have been (6), with the overt complementizer that in the very position that we are assuming a silent complementizer in the modern language.
(6) | They wonder which wildebeest that the lions will devour. |
Such sequences of wh-phrase + overt complementizer are also attested in a number of other modern languages (mostly in vernacular varieties). Indeed, even in modern English, examples structurally parallel to (6) are heard from time to time in unedited usage. The examples in (7) are ones that I have encountered (or produced myself), and further examples are given in Radford 1988:500.
(7) | a. | I'm just trying to figure out how that it's all gonna be
facilitated.
(Quin Bradley Lee, 19 November 1996; how with pitch accent) | |
b. | Immediately, I saw which one that you wanted me to read.
(Beatrice Santorini, September 1998) | ||
c. | Are you more interested where in the mouth that
the sounds are produced?
(Beatrice Santorini, 9 December 1998) | ||
d. |
it is relatively easy to tell where along the path
of the formation of the utterance that an error occurred.
(Marya Grzesiak, term paper draft, 14 December 1998) |
(8) | She thinks he/*him will come. |
Now consider (9), where the nominative form who is grammatical, but the oblique form whom is not.
(9) | Who/*Whom does she think will come? |
We would like to maintain the generalization that nominative case is assigned under spec-head agreement with finite I. But if we were to let who receive nominative case from the finite I of the matrix clause (however that would be implemented), then she would have no source of nominative case. Again, the solution to the puzzle is to invoke movement. If we assign (9) the structure in (10), then who can receive nominative case from the finite I of the complement clause, and she receives its nominative case from the matrix finite I, as expected.
(10) | Whoi/*Whomi does she think [IP ti will come ] ? |
The movement approach embodied in (10) is supported over conceivable alternatives by the parallel in (11).
(11) | a. | Whom does she expect to come? | |
b. | She expects him/*he to come. |
Note: (12) omits some structural details of movement to C that are irrelevant for present purposes. The details in question are covered in Notes 11.
(12) |
It is important to understand that movement to C is independent of movement to Spec(CP). The table in (13) shows how the two movements combine to yield various types of clauses.
(13)
Movement to C? | |||
---|---|---|---|
Yes | No | ||
Movement to Spec(CP)? | Yes | Direct wh-questions | Indirect wh-questions |
No | Direct yes-no questions | Other |
(14) | a. | [ Which book ]i was he reading ti? | |
b. | [ Which book ]i did he say that he was reading ti? | ||
c. | [ Which book ]i did she believe that he said that he was reading ti? | ||
d. | [ Which book ]i did they claim that she believed that he said that he was reading ti? | ||
e. | [ Which book ]i do you think they claimed that she believed that he said that he was reading ti? |
The simplest way to account for the facts in (14) is to assume that a wh-phrase can move from its original position to a sentence-initial Spec(CP) position in one fell swoop. This is illustrated for (14b) in (15).
(15) |
An alternative is that the movement takes place in two hops. The first hop takes the moved constituent from its original position to the Spec(CP) of the complement clause, which functions as an intermediate landing site. The second hop takes it from there to its destination, the Spec(CP) of the matrix clause. The difference between the one-fell-swoop derivation in (15) and its alternative in (16) is the presence of the intermediate trace, indicated by a box. Notice that in (16), the moved constituent never leaves more than a single IP on any one hop. This will become important directly.
(16) |
(17) | a. | They wonder how quickly the lions will devour the wildebeest. | |
b. | * | [ Which wildebeest ]i do they wonder how quickly the lions will devour ti? |
The tree for (17b) is given in (18).
(18) |
If wh-movement were able to occur in one fell swoop, then the movement represented in (18) should be possible. But if we assume the second alternative, according to which apparent long-distance wh-movement is actually composed of several short-distance hops, then the ungrammaticality of (17b) is not surprising.
In particular, we could say that IP nodes form barriers to movement (indicated by boxes in (18)), and that not more than one such barrier may be crossed on any one hop. This proposed property is called subjacency. Subjacency means that a wh-constituent can escape out of its IP if there is a local (= neighboring) Spec(CP) that is empty and available as an intermediate landing site, as is the case in (16). But if the local Spec(CP) is not available because it is already occupied, the wh-constituent is 'trapped' and is unable to move.
IP nodes are not the only barriers in English. DP nodes belong to the set of barriers as well. We know this because if we try to move a wh-constituent out of an DP node and an IP node on one hop, the result is ungrammatical. An example of attempted movement out of a DP and an IP node is shown in (19b), with a structure as in (20).
(19) | a. | He should buy the picture with the gold frame. | |
b. | * | [ Which frame ]i should he buy the picture with ti? |
(20) |